ChatGPT Checker

The ChatGPT Checker analyzes a prompt and its corresponding model output against a configurable rubric that covers factual accuracy, completeness, bias indicators, and citation quality. It is designed for AI QA engineers, researchers, and prompt designers who need repeatable, auditable assessments across model iterations. The tool does not modify or execute the model; it evaluates outputs against supplied criteria and references to guide improvement.
Conceptually, it ingests inputs, applies a scoring logic, and returns structured results, including per-criterion scores, an overall score, and flagged issues with explanations. It supports configurable weights, optional reference facts, and exportable reports for traceability.
Unique value lies in configurable rubrics, reproducible scoring across versions, and claim-source mapping to references. Use cases include prompt testing, model benchmarking, and regulatory-grade evaluation artifacts.
Core features include factuality checks against references, missing citation detection, bias and toxicity cues (optional), language handling, and export formats suitable for QA pipelines. Outputs enable data-driven prompt refinement and evidence-backed model documentation.